Jump to content

Talk:Tesla Dojo

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia


Should this be a standalone article?

[edit]
The following discussion is closed. Please do not modify it. Subsequent comments should be made in a new section. A summary of the conclusions reached follows.
Closing and proceeding with merge. There have been no recent updates, and multiple editors agree this is clear case of WP:CRYSTALBALL, which says: Until such time that more encyclopedic knowledge about the product can be verified, product announcements should be merged to a larger topic...if applicable. Stonkaments (talk) 22:19, 17 October 2021 (UTC)[reply]

I don't think Tesla Dojo is notable enough to warrant a standalone article. I propose merging into Tesla Autopilot (or maybe History of Tesla, Inc.) instead, at least until we wait and see if the project ultimately gains notability over time. The chip is yet to be released, and all of the sources are simply echoing Tesla's PR announcement yesterday. And of course Tesla has announced numerous projects over the years that have failed to come to fruition: plans to produce a COVID-19 vaccine[1], battery swapping[2], robotaxis[3], Model S Plaid Plus[4], etc. WP:NOTNEWS says "routine news reporting of announcements, sports, or celebrities is not a sufficient basis for inclusion in the encyclopedia." Stonkaments (talk) 11:34, 20 August 2021 (UTC)[reply]

There is an Fugaku (supercomputer) and all of the top 8 also have articles. If this is expected to get to an exaflop soon (was this year claimed to be expected?) which is twice the flop rating of the current top supercomputer as at June 2021 then it would deserve an article just on that basis. Don't think I would be against a Tesla AI Day article. C-randles (talk) 17:48, 20 August 2021 (UTC)[reply]
I don't think Fugaku is a good comparison; that is an actual supercomputer that has been launched, and has received significant media coverage beyond the initial PR announcement. Tesla Dojo is just another one of Musk's crazy promises[5]; the specs they announced have no bearing on what the end result will actually be. Stonkaments (talk) 20:24, 20 August 2021 (UTC)[reply]
KEEP Added a link precursor GPU based computer which says "believe that in terms of flops this is roughly the number five supercomputer in the world” If the precursor existing supercomputer is up there already at a level where we seem to have articles then it's successor may be of interest. In addition, this has multiple potential interested audiences - supercomputers, tesla, self driving cars, AI so I think it should be kept. C-randles (talk) 10:52, 21 August 2021 (UTC)[reply]
When the specifications change, those changes should be recorded. That they are malleable and subject to change has no impact on the notability of the subject. --No coffee, please. (talk) 23:31, 21 September 2021 (UTC)[reply]
Merge as the "supercomputer" is part of the larger Autopilot/FSD project. QRep2020 (talk) 16:32, 21 August 2021 (UTC)[reply]
  • KEEP I think this should be a standalone article and should not be merged to Tesla, Inc. or Tesla Autopilot as it is supposed to be a new chip (hardware) technology like CPU, GPU or even Tensor Processing Unit are. We are facing since few years a deep revolution of the hardware components and others company should have no problems to implement a similar technology in their products. More over it is clear that Tesla will not keep this technology as an inclusive service. I think it is safer to create a standalone page for DPU as the D1 is the technology chip created and named by Tesla to describe a new technology architecture, the DPU. Two appart things. We should consider merging Tesla Dojo in the more precise draft: DPU / WP:DRAFT (Tirbo06 (talk) 09:59, 27 August 2021 (UTC)) Tirbo06 (talkcontribs) has made few or no other edits outside this topic. [reply]
  • Merge into Tesla Autopilot. The content is a stub at this point. Tesla is able to capture enough of the news cycle so it's easy to find hits but depth is lacking. Since the whole system is autopilot focused it should live there. The view that this is something new may be valid but we should wait until it proves to be true (see CYRSTAL). Tesla has a very long history of over promise then under/never deliver. Springee (talk) 12:27, 31 August 2021 (UTC)[reply]
  • Keep, I've created this article and I can admit that it was too early, but this was an interesting topic and an interesting product, that I think is notable by itself, not just as part of Autopilot. Though, if it will be just an announcement with no further action, it can be merged into History of Tesla (if such an article exist, or to the Autopilot article later). Artem.G (talk) 18:35, 5 September 2021 (UTC)[reply]
Isn't that effectively CRISTAL logic? If this turns out to be more that the typical Tesla hype... ETA TBD... shouldn't this be merged first rather than later? I'm not sure this RfC will answer much since this question would normally be answered at AfD vs on the article talk page. Springee (talk) 23:37, 6 September 2021 (UTC)[reply]
Yeah, I wasn't sure if AfD or a merger proposal would make more sense. I just pinged some relevant WikiProjects, so hopefully we'll get some more input from the broader community shortly. Stonkaments (talk) 22:06, 21 September 2021 (UTC)[reply]
  • Keep to me, and a strong one. Fundamentally, the likelihood that a project will come to fruition or not should not influence notability, and Dojo is definitely fundamentally a different project from Autopilot. I think you're really reaching with WP:NOTNEWS here. Worth pointing out that I understand the "just another one of Musk's crazy promises" argument, but there's strong evidence this isn't the case here. Tesla has an established history of developing hardware with HW3.0, and Musk has been discussing Dojo for a long time, not merely at AI Day. A better use of energy might be keeping a watch on language being used on the page, since there's certainly a history of over-promising and under-delivering. --No coffee, please. (talk) 23:27, 21 September 2021 (UTC)[reply]
  • Merge - These discussions tend to attract a disproportionate amount of Tesla aficionados (just look at the size of Tesla, Inc.) so this is going to be the unpopular opinion, but there is clear Wikipedia policy that speaks directly to this issue in particular, WP:CRYSTALBALL.

    5. Wikipedia is not a collection of product announcements and rumors. Although Wikipedia includes up-to-date knowledge about newly revealed products, short articles that consist of only product announcement information and rumors are not appropriate. Until such time that more encyclopedic knowledge about the product can be verified, product announcements should be merged to a larger topic (such as an article about the creator(s), a series of products, or a previous product) if applicable.

    Until such time as there is verifiable information it should be part of another article. IPBilly (talk) 00:13, 22 September 2021 (UTC)[reply]
Nah, I don't see it. The key phrase here is ``"short articles that consist of only product announcement information and rumours are not appropriate"```. However, we're past the announcement stage here. As the article says, the announcements started back in 2019 and 2020. Musk has been tweeting about it ever since. That's when it would have been inappropriate to create a new page. We're now at the point of architectural layouts, detailed images of the intended product, detailed descriptions of what they're trying to achieve, and very detailed target specifications. Whether you believe that the announced specs will be the final specs at all (i personally do not) is irrelevant. The point is to begin collecting encyclopedic information about the topic as it develops, and note the changes along the way. I'd say we're like riiiiight at the point of that since "AI Day". If you want to pick a similar product that is illustrative of being at the announcement stage and qualifies for WP:CRYSTALBALL, the Tesla Bot is a perfect one. --No coffee, please. (talk) 02:00, 22 September 2021 (UTC)e[reply]
I respectfully disagree. 1) short articles: There's already been more written here than the entire article, which is a scant 12 sentences. 2) A couple of cryptic (per the cited source) tweets from Elon hardly count as hyperbole, let alone the start of announcements. 3) Are we past the announcement stage? As you say, there's only images of the intended product, descriptions of what they're trying to achieve, and target specifications. I realize claims in this field can only really be taken at face value as they're hard to independently verify, but all these sources are simply rehashing what Tesla has published and speculating on the rest. Without more it's a hard sell to me. IPBilly (talk) 04:26, 22 September 2021 (UTC)[reply]
Keep in mind, this isn't a consumer product. You're not really going to get more. Anandtech is not going to do a hands-on. No one is going to be given one for a week to play around with. Most information will come from primary sources and benchmarks run by Tesla themselves, and that will be the likely battle this article faces as it evolves. This article is going to parallel Tensor Processing Unit or Summit (supercomputer) in overall depth of information, that's simply the likely shape it will take.
I use the phrases intended product, trying to achieve, and target specifications because that's the most intellectually honest framing I can use, and because I personally doubt all of those things will stay static. Tesla's presentation is of course, more sure-footed. Neither framing should have any relevance over the notability of the topic. --No coffee, please. (talk) 05:27, 22 September 2021 (UTC)[reply]
The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

So Stonkaments opened this discussion, got 3 people saying merge (plus him/herself?) and 6 people saying keep. the crystalball argument for merge was disputed and Stonkaments closes it as merge. Hmm. C-randles (talk) 15:17, 18 October 2021 (UTC)[reply]

Yeah, agree this is rather shitty behaviour. Pretty clear what Stonkaments is doing here, given how clear the votes were. --No coffee, please. (talk) 04:18, 19 October 2021 (UTC)[reply]

article of interest

[edit]

The Tesla Dojo Chip Is Impressive, But There Are Some Major Technical Issues, semianalysis, [[6]] Springee (talk) 12:39, 31 August 2021 (UTC)[reply]

Added C-randles (talk) 17:15, 1 September 2021 (UTC)[reply]

Well... Dojo is automated, so I guess we can try it.

Factual errors in "The Next Platform" article used as reference.

[edit]

Reference 5 (Morgan, Timothy Prickett (August 23, 2022). "Inside Tesla's innovative and homegrown 'Dojo' AI supercomputer". The Next Platform. Retrieved 12 April 2023.) in this article contains numerous factual errors, including misinterpreting bytes as bits, and some other issues which I have corrected in my latest edit.

I noticed that this source is used across a lot of this wiki article, and indeed there are some sections that appear to have been plagiarised entirely from that source.

To ensure accuracy of information, this source should probably be removed and prose about specific architectural details should be rewritten, citing reference 27 ("The Microarchitecture of DOJO, Tesla's Exa-Scale Computer". IEEE Micro. 43 (3): 31–39.) instead if needed.

I may not have caught all the incorrect information, so any help would be appreciated. - B4shful (talk) 19:12, 12 December 2023 (UTC)[reply]